A Complete Guide to the AI Chatbot App

Chinese AI lab DeepSeek gained widespread attention this week as its chatbot app surged to the top of both the Apple App Store and Google Play charts. The company’s AI models, developed with compute-efficient methods, have prompted Wall Street analysts and tech experts to question the U.S.’s ability to maintain its leadership in AI and whether the demand for AI chips will remain strong.

So, what are DeepSeek’s origins, and how did it achieve global recognition so rapidly?

DeepSeek is supported by High-Flyer Capital Management, a Chinese quantitative hedge fund that leverages AI for its trading strategies.

From Student Trader to AI-Driven Hedge Fund Founder

AI enthusiast Liang Wenfeng co-founded High-Flyer in 2015. Having started exploring trading as a student at Zhejiang University, Wenfeng established High-Flyer Capital Management as a hedge fund in 2019, focusing on creating and implementing AI algorithms.

In 2023, High-Flyer launched DeepSeek as a separate lab dedicated to AI research, distinct from its financial operations. With High-Flyer as an investor, DeepSeek eventually became an independent company under the same name.

Building Infrastructure Amid U.S. Hardware Export Restrictions

From the beginning, DeepSeek developed its own data center clusters for model training. However, like many Chinese AI firms, it has faced challenges due to U.S. export restrictions on hardware. For training one of its latest models, DeepSeek had to rely on Nvidia H800 chips—a less powerful alternative to the H100 chips available to U.S. companies.

DeepSeek’s technical team is reportedly quite young. The company is known for actively recruiting PhD-level AI researchers from leading Chinese universities. Additionally, DeepSeek hires individuals without computer science backgrounds to help its technology gain a broader understanding of various topics, according to The New York Times.

DeepSeek introduced its initial models—DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat—in November 2023. However, it wasn’t until last spring, with the release of its next-generation DeepSeek-V2 models, that the AI community began to take serious notice.

DeepSeek-V2, a versatile system for analyzing text and images, performed strongly across various AI benchmarks and was significantly more cost-effective to operate than comparable models at the time. This pressured domestic competitors like ByteDance and Alibaba to lower prices on some of their models and offer others for free.

The launch of DeepSeek-V3 in December 2024 further boosted the company’s reputation.

DeepSeek V3 Outperforms Leading Open and Closed AI Models

According to internal tests, DeepSeek V3 surpasses both downloadable open models like Meta’s Llama and closed API-only models such as OpenAI’s GPT-4o.

Another standout is DeepSeek’s R1 “reasoning” model, released in January, which DeepSeek claims matches the performance of OpenAI’s o1 model on key benchmarks.

As a reasoning model, R1 can effectively fact-check itself, helping it avoid common errors that typically challenge AI models. Although reasoning models take longer—usually seconds to minutes more—to reach conclusions compared to standard models, they offer greater reliability in fields like physics, science, and math.

Regulatory Restrictions Limit DeepSeek’s AI Responses

There is a drawback to R1, DeepSeek V3, and the company’s other models. As Chinese-developed AI, they undergo evaluation by China’s internet regulator to ensure their responses align with “core socialist values.” For instance, DeepSeek’s chatbot won’t address questions about Tiananmen Square or Taiwan’s autonomy.

In March, DeepSeek recorded over 16.5 million visits. “[F]or March, DeepSeek ranks second, despite a 25% drop in traffic compared to February, based on daily visits,” David Carr, editor at Similarweb, told TechCrunch. However, this is still far behind ChatGPT, which surpassed 500 million weekly active users in March.

In May, DeepSeek released an updated version of its R1 reasoning AI model on the developer platform Hugging Face.

If DeepSeek has a business model, it’s not entirely clear what it is. The company offers its products and services at prices well below market rates—and even provides some for free. Despite significant interest from venture capitalists, DeepSeek is not currently accepting investor funding.

Efficiency Claims Drive Low Costs, but Experts Remain Skeptical

DeepSeek claims that breakthroughs in efficiency allow it to keep costs extremely low, though some experts question the accuracy of these claims.

Regardless, developers have embraced DeepSeek’s models. While not open source in the traditional sense, they are available under permissive licenses that permit commercial use. Clem Delangue, CEO of Hugging Face—a platform hosting DeepSeek’s models—reported that developers have created over 500 “derivative” models based on R1, collectively downloaded 2.5 million times.

DeepSeek’s rapid rise against larger, established competitors has been described as “upending AI” by some and “overhyped” by others. Its success contributed to an 18% drop in Nvidia’s stock price in January and prompted a public response from OpenAI CEO Sam Altman. According to Reuters, U.S. Commerce Department agencies announced in March that DeepSeek would be banned on government devices.

Integration, Investment, and Controversy Surrounding DeepSeek

Microsoft has integrated DeepSeek into its Azure AI Foundry service, which consolidates AI services for enterprises. When asked about DeepSeek’s effect on Meta’s AI investments during a first-quarter earnings call, CEO Mark Zuckerberg affirmed that AI infrastructure spending remains a “strategic advantage” for Meta. Meanwhile, in March, OpenAI labeled DeepSeek as “state-subsidized” and “state-controlled,” recommending that the U.S. government consider banning its models.

During Nvidia’s fourth-quarter earnings call, CEO Jensen Huang praised DeepSeek’s “excellent innovation,” noting that reasoning models like DeepSeek’s require significant computing power, benefiting Nvidia.

At the same time, some organizations, countries, and governments—including South Korea and New York State—have banned DeepSeek on official devices. In May, Microsoft Vice Chairman and President Brad Smith testified before the Senate that Microsoft employees are prohibited from using DeepSeek due to concerns over data security and propaganda.

As for DeepSeek’s future, it remains uncertain. Improved models are expected, but the U.S. government appears increasingly cautious about potential foreign influence. The Wall Street Journal reported in March that the U.S. will likely ban DeepSeek on government devices.

Read the original article on: Techcrunch

Read more: Google and Duolingo think AI can transform language learning. Do they?